Why SPC Belongs in Deep Hole Drilling
Statistical process control transforms deep hole drilling from a reactive process to a predictive one. Instead of finding out a part is bad when you measure it after the cycle, SPC tells you the process is trending toward a problem before the part is made.
I started using SPC on a hydraulic cylinder job that was running 500 parts per month. The rejection rate was 3% from bore diameter drift. The problem was not that the process was unstable — it was that I could not see the drift until I measured the finished part.
SPC changed that. By measuring every 5th part and plotting the results, I could see the diameter trending upward as the tool wore. I changed the tool based on the chart, not based on the inspection results. The rejection rate dropped to under 0.5%.
The investment in SPC is small. A notebook and a calculator are enough to get started. A digital spreadsheet or dedicated SPC software makes it easier, but the core method works with pencil and paper.
What to Measure and How Often
Key Variables
I track three variables on every production deep hole drilling job. The choice of variables depends on what the print calls out, but these three cover most applications.
| Variable | Chart Type | Measurement Frequency |
|---|---|---|
| Hole diameter | X-bar (average) | Every 5th part |
| Diameter range | R (range) | Every 5th part |
| Surface finish | Individual | Every 10th part |
Hole diameter is the primary variable because it is the most common critical dimension on a deep hole print. I measure with an air gauge or a bore micrometer to 0.001mm resolution.
Diameter range — the difference between the maximum and minimum diameter within a single part — catches ovality and taper. I measure at three depths in each hole: near entry, mid-depth, and near bottom.
Sampling Frequency
The sampling frequency depends on the production volume and the process stability. For a stable process running high volume, every 5th part provides enough data. For an unstable or new process, I measure every part until I understand the behavior.
| Production Volume | Sampling Frequency |
|---|---|
| Under 50 parts | Every part |
| 50-200 parts | Every 3rd part |
| 200-1000 parts | Every 5th part |
| Over 1000 parts | Every 10th part |
I adjust the frequency based on the chart data. If the X-bar chart starts showing more variation, I increase the sampling frequency until the process stabilizes.
Building and Reading Control Charts
Setting Up the Chart
I collect 20 initial samples to establish the baseline. For a job measuring every 5th part, that means the first 100 parts provide the control limits. I calculate the average (X-bar) and range (R) for each subgroup.
The upper control limit (UCL) and lower control limit (LCL) are set at plus and minus three sigma from the process average. These limits define the range of normal process variation.
| Control Limit | Calculation | Meaning |
|---|---|---|
| UCL | Average + 3 sigma | If exceeded, process is out of control |
| Average | Mean of all subgroup averages | Target for the process |
| LCL | Average - 3 sigma | If exceeded, process is out of control |
What to Look For
I look for four patterns on the control chart. Each pattern indicates a different type of process change.
| Pattern | Indication | Action |
|---|---|---|
| Point above UCL or below LCL | Process out of control | Stop and investigate immediately |
| 7 consecutive points above or below average | Process shift | Check for cause of shift |
| 7 consecutive points trending up or down | Tool wear or drift | Adjust or change tool |
| Increasing range | Process becoming unstable | Check coolant, material, setup |
An example from my experience. The bore diameter on a cylinder job was trending upward slowly over 50 parts. Each individual measurement was within tolerance, but the trend was clear on the chart. I changed the tool and the diameter returned to nominal. The trend caught the wear before it caused a reject.
Making SPC Practical on the Shop Floor
Data Collection
I keep data collection sheets at each machine. The operator measures the part and writes the reading on the sheet. The sheet has pre-calculated control limits so the operator can see immediately if a reading is out of range.
The operator does not need to do the math. The sheet has a visual reference — a line for the UCL, a line for the LCL, and the target in between. The operator plots the reading on the chart and knows right away if something is wrong.
Response Protocol
When a reading falls outside the control limits, I have a standard response protocol. The operator stops production and calls me. I investigate before any more parts are run.
| Out-of-Control Signal | Immediate Action |
|---|---|
| Point above UCL | Stop, measure tool, check coolant |
| Point below LCL | Stop, check material, verify gauge |
| Run of 7 above average | Prepare for tool change |
| Run of 7 below average | Check material batch change |
Long-Term Benefits
After running SPC on a job for several months, I have a detailed history of the process behavior. That history helps me set realistic control limits, plan tool changes, and predict maintenance needs.
I have used SPC data to justify changing tool grades, adjusting coolant concentration, and modifying feed rates. The data does not lie — if the chart shows a persistent problem, the process needs a change.
Key Takeaways
- SPC catches process drift before it produces scrap — it is predictive, not reactive.
- Track hole diameter on X-bar and R charts with measurements every 5th part.
- Set control limits at plus and minus three sigma from the process average.
- Look for trends, shifts, and out-of-control points on the chart.
- Keep data collection sheets at the machine with visual control limit references.
- Use SPC history to justify process changes and optimize tool change intervals.